Neuron Explorer

Qwen3-0.6B

opening parenthesis before a numbered list

A neuron in layer 16 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on an opening bracket '(' that starts a numbered list, such as '(1)' after a line like 'four critical areas:'. It is quiet, or even pushed the other way, on brackets that follow a term or a quoted name. Turning it up caused repeated word patterns rather than clear list text.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    ability to apply those principles in sermon preparation, personal Bible study, or writing. This popular book outlines a five-step hermeneutical procedure that includes: (1) historical-cultural and contextual analysis, (2 /5(89).

  2. #2 peak

    who<|endoftext|>OF FATTY ACID METABOLISM⏎coma due to a disorder of fatty acid oxidation may occur with any of the⏎following conditions: (1) deficiency of short-chain, medium-chain, or⏎long-chain acyl

  3. #3 peak

    time should focus on four critical areas: (1) developing understanding of and applying proportional relationships; (2) developing understanding of

  4. #4 peak

    wish to perpetuate naturalism and Darwinism in our culture (the "National Standards").⏎The National Standards promote the philosophy of naturalism in three fundamental ways: (1) by incorporating the philosophy of naturalism into the definition of science, (

  5. #5 peak

    as part of its regional initiative, Indigenous Voices: Communication for Empowerment of Asia’s Indigenous Peoples. Each pilot initiative consisted of three core elements: (i) review of the media context based on the existing research at the national level

  6. #6 peak

    This preview has intentionally blurred sections. Sign up to view the full version.View Full Document⏎Unformatted text preview: itself, and if ( ) ( ) lim lim x c x c h x L g x = =

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: (a particular (elementary problem: ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( - ( - ( - - ( - ( - - ( - - - - - ( - - - - -

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    and sacred sites were designed to mark solar and astrological events. A stone, Intihuatana ('Hitching Post of the Sun'), is the focus of a major religious site

  2. #2 peak

    Gate became the site of the Indian Army's Tomb of the Unknown Soldier, known as Amar Jawan Jyoti ("the flame of the immortal soldier").⏎Until the 1920s,

  3. #3 peak

    about sex, about rest.” Evelyn, a 34-year-old Dominican immigrant who recently gave birth, is explaining the Latin American custom called la cuarentena (“quarantine”). It’s a 40-day postpartum period during

  4. #4 peak

    hold traditional territories in the region and own and manage the remaining 2.7% of the Watershed.⏎The Yukon has several First Nation Final Agreements (“Final Agreements”) with various Yukon First Nations. These Final Agreements adopt

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L5:330 excites opening parenthesis after a noun
  2. L9:339 excites
  3. L8:354 excites opening parenthesis before a citation year
  4. L10:66 suppresses opening parenthesis after a citation or term
  5. L14:90 suppresses
  6. L14:1209 excites
  7. L10:142 suppresses opening parenthesis mid-sentence
  8. L3:57 excites
  9. L10:1459 suppresses
  10. L10:1041 suppresses opening parenthesis before phone or code
  11. L4:45 suppresses
  12. L5:59 suppresses
  13. L6:144 excites
  14. L14:2402 weakly suppresses
  15. L11:2191 weakly excites unclear: brackets and parentheses in text
  16. L4:1088 weakly suppresses
  17. L4:64 weakly excites
  18. L15:381 weakly suppresses
  19. L5:1061 weakly suppresses
  20. L9:2300 weakly excites
  21. L11:50 weakly excites
  22. L0:86 weakly excites
  23. L15:123 weakly suppresses
  24. L8:2208 weakly suppresses

Feeds later layers

  1. L19:2251 excites opening parenthesis after a name
  2. L18:980 excites opening parenthesis after a noun phrase
  3. L18:1501 suppresses
  4. L17:2018 suppresses
  5. L18:682 excites
  6. L18:2337 suppresses
  7. L17:1509 excites
  8. L23:1619 excites
  9. L23:224 weakly suppresses
  10. L20:1451 weakly excites
  11. L25:360 weakly suppresses
  12. L26:2786 weakly suppresses
  13. L17:3049 weakly excites
  14. L17:137 weakly suppresses
  15. L22:3047 weakly excites
  16. L17:2233 weakly suppresses 'any' after negatives or 'more'
  17. L18:2092 weakly suppresses
  18. L17:364 weakly suppresses hyphen joining two words
  19. L17:2365 weakly excites
  20. L19:2614 weakly excites
  21. L18:762 weakly suppresses
  22. L19:453 weakly excites
  23. L20:104 weakly suppresses
  24. L24:2099 weakly excites

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -4.24 and -3.921 values between -3.60 and -3.283 values between -3.28 and -2.961 values between -2.96 and -2.642 values between -2.64 and -2.322 values between -2.32 and -2.008 values between -2.00 and -1.687 values between -1.68 and -1.3620 values between -1.36 and -1.0433 values between -1.04 and -0.7262 values between -0.72 and -0.40429 values between -0.40 and -0.082485 values between -0.08 and 0.24751 values between 0.24 and 0.5667 values between 0.56 and 0.8815 values between 0.88 and 1.207 values between 1.20 and 1.525 values between 1.52 and 1.842 values between 3.12 and 3.441 values between 3.44 and 3.761 values between 6.00 and 6.322 values between 6.64 and 6.961 values between 7.28 and 7.601 values between 7.60 and 7.921 values between 8.24 and 8.56offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. 以下简称
  2. Directive
  3. 就是
  4. 也就是
  5. including
  6. 即
  7. 바로
  8. co
  9. Ctrl
  10. )</

Suppresses

  1. makeStyles
  2. arians
  3. -ul
  4. .Profile
  5. abis
  6. epend
  7. erta
  8. arian
  9. abad
  10. 玒